error when running Real ESRGAN x4 models with x2 scale

[ssube]

While not strictly supported, this should not crash the job:

[2023-02-16 01:00:16,333] INFO: onnx_web.chain.upscale_resrgan: upscaling image with Real ESRGAN: x4                                                                      
EP Error using [('CUDAExecutionProvider', {'device_id': 0})]
Falling back to ['CUDAExecutionProvider', 'CPUExecutionProvider'] and retrying.
[2023-02-16 01:00:16,489] INFO: werkzeug: 10.2.2.16 - - [16/Feb/2023 01:00:16] "GET /api/ready?output=upscale_1434952017_1641dbbc27dade5d93110f18e956e44b6707f773dff34380b5a64976d570e7b2_1676509216.png HTTP/1.1" 200 -                                                                                                                                              
[2023-02-16 01:00:21,541] INFO: werkzeug: 10.2.2.16 - - [16/Feb/2023 01:00:21] "GET /api/ready?output=upscale_1434952017_1641dbbc27dade5d93110f18e956e44b6707f773dff34380b5a64976d570e7b2_1676509216.png HTTP/1.1" 200 -                                                                                                                                                
        Tile 1/1                                                                                                                                                                    
[2023-02-16 01:00:21,687] WARNING: onnx_web.server.device_pool: job upscale_1434952017_1641dbbc27dade5d93110f18e956e44b6707f773dff34380b5a64976d570e7b2_1676509216.png failed with an error: ['

Traceback (most recent call last):
', '  File "/home/ssube/onnx-web/api/onnx_web/server/device_pool.py", line 243, in job_done
    f.result()
', '  File "/usr/lib/python3.10/concurrent/futures/_base.py", line 451, in result
    return self.__get_result()
', '  File "/usr/lib/python3.10/concurrent/futures/_base.py", line 403, in __get_result
    raise self._exception
', '  File "/usr/lib/python3.10/concurrent/futures/thread.py", line 58, in run
    result = self.fn(*self.args, **self.kwargs)
', '  File "/home/ssube/onnx-web/api/onnx_web/diffusion/run.py", line 221, in run_upscale_pipeline
    image = run_upscale_correction(
', '  File "/home/ssube/onnx-web/api/onnx_web/server/upscale.py", line 61, in run_upscale_correction
    return chain(
', '  File "/home/ssube/onnx-web/api/onnx_web/chain/base.py", line 147, in __call__
    image = stage_pipe(
', '  File "/home/ssube/onnx-web/api/onnx_web/chain/upscale_resrgan.py", line 112, in upscale_resrgan
    output, _ = upsampler.enhance(output, outscale=upscale.outscale)
', '  File "/home/ssube/onnx-web/api/onnx_env/lib/python3.10/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
', '  File "/home/ssube/onnx-web/api/onnx_env/lib/python3.10/site-packages/realesrgan/utils.py", line 221, in enhance
    self.tile_process()
', '  File "/home/ssube/onnx-web/api/onnx_env/lib/python3.10/site-packages/realesrgan/utils.py", line 178, in tile_process
    self.output[:, :, output_start_y:output_end_y,
', 'RuntimeError: The expanded size of the tensor (2048) must match the existing size (1024) at non-singleton dimension 3.  Target sizes: [1, 3, 2048, 2048].  Tensor sizes: [3, 1024, 1024]

']

[ssube]

This does not occur in txt2img mode, and seems to be working correctly in upscaling mode for my test inputs:

• 512x512
• 256x256
• 1024x768

It does end up being cut off, because of the mismatch between model scale and parameters scale/outscale: